agents on the live web benchmarking information extraction when the ground truth keeps moving speech & audio language models pre-processing, evaluation, and artifacts in full-duplex speech systems retrieval & multilingual alignment finding the right context, and closing the gap between languages